Cursor (@cursor_ai) launched two last-mile bots: Rollouts tracks a change from PR open through production verification—writing a monitoring plan, comparing post-deploy signals to baseline, and optionally pinging the author, pausing a progressive rollout, or opening a revert PR; Security Reviewer now averages 3.8 minutes per review (down from 4.8, ~21% faster). Both are live on Teams/Enterprise, with 10 days of Rollouts usage credits.

Key Takeaways

  • ✓Official blog: Rollouts connects SCM, deploy, and Datadog/Grafana/Honeycomb telemetry; writes a pre-merge monitoring plan and diffs post-deploy signals vs baseline.
  • ✓Security Reviewer averages 3.8 min (was 4.8) — ~**21%** faster; findings include severity, attack path, and one-click fix.
  • ✓Catches single-endpoint/region regressions, separates intended effects from real regressions, and flags missing instrumentation before merge.
  • ✓Available today on Teams/Enterprise with 10 days of Rollouts credits; enable from the Automations tab.
🔗

Project Links & Resources

Direct Access
Direct access to official project resources and documentation
📝
Bots for the last mile: Rollouts, Security Review
cursor.com/blog/rollouts-and-security-reviewer
↗
📖
Cursor Docs
docs.cursor.com
↗
🔬

In-Depth Technical Analysis

Core Background & Industry Pain Points Agents sped up coding; post-PR security review, deploy watching, and regression attribution did not. Cursor’s last-mile bots post targets that repetitive, context-heavy work. ### Architecture Highlights & Internals Rollouts writes a pre-merge monitoring plan from the diff, then compares post-deploy signals to baseline; on regression it can notify, pause progressive rollout, or open a revert PR. Security Reviewer reads each PR in whole-repo context and follows input→sink dataflow for injection, auth gaps, secrets, unsafe deserialization, vulnerable deps, and insecure defaults. ### Authoritative Benchmarks & Measured Scores Security Reviewer avg 3.8 min (was 4.8), ~21% faster. Rollouts has no public SWE leaderboard; Cursor highlights single-endpoint/region catch, intended-vs-regression separation, and missing-instrumentation flags. ### Developer Hands-on Guide Enable on Teams/Enterprise Automations; wire SCM, deploy, and Datadog/Grafana/Honeycomb. Credits for Rollouts for 10 days. Blog: cursor.com/blog/rollouts-and-security-reviewer.

⚡
Evaluating this AI coding model or solution?
Check live multi-benchmark rankings or compare plan costs & promo credits.
🏆 Leaderboard→💰 Compare Plans↗
ADSponsored